The Mars 2020 Rover will likely contain Artificial Intelligence software to handle the scientific workload.  Welcome to Ars UNITE, our week-long virtual conference on how innovations bring together unusual pairings. Every day of the week from Wednesday to Friday we bring you some stories about the future. The focus today is on AI in manufacturing and space ̵
1; just wait to get started!
NASA can not bring scientists to Mars yet. NASA's Jet Propulsion Laboratory hopes to use artificial intelligence on its next rover mission on the Red Planet, to use at least the equivalent of a talented science assistant. Steve Chien, head of the AI group at NASA JPL, envisioned the work on the Mars 2020 Rover "more like [how] you'd interact with a doctoral student, rather than a rover you usually need to micromanagement."
The 13-minute delay in Earth-Mars communications means that the movements and experiments performed by previous and current Mars rovers must be carefully planned. While newer Rovers were already able to detect threats and perform some tasks autonomously, they still placed high demands on their support teams.
Chien sees a role in the future role of AI in space, in which people focus on hard parts, such as controlling robots naturally, while machines work autonomously and give people a high-level summary give.
"The AI will almost be like a partner with us," Chien predicted. "It will try, and then we'll say, 'No, try something that's longer, because I think that might look better,' and then it'll try. It understands what's elongated, and it knows much of the details like trying to fly the formations, that's the next level.
"Then, of course, it becomes tangible on the dystopian level," quipped Chien, but he does not see it that fast.
Old – school autonomy
NASA has a long history of AI and machine learning technologies, Chien said, much of which has focused on using machine learning to interpret extremely large amounts of data since the data is sent back to Earth for processing There is a good reason to integrate more intelligence directly into the spacecraft: to support communication.
Earth Observing One was an early example of introducing Informatio on board a spacecraft. Launched in November 2000, EO-1 was originally scheduled to have a one-year mission to test, among other things, how basic AI could perform some scientific tasks on board. One of the AI systems tested on board EO-1 was the Autonomous Sciencecraft Experiment (ASE), a set of software that allowed the satellite to make decisions based on the data collected by its image sensors. ASE included integrated scientific algorithms that performed image data analysis to detect trigger conditions, so that spacecraft should pay more attention to something, such as discovered interesting features or changes from previous observations. The software could also detect Cloud Cover and edit it from the final home transferred image packages. The EO-1 ASE could also adapt the satellite's activities based on the findings collected in an earlier orbit.
Using volcanic images, Chien said that JPL had trained the machine learning software to detect volcanic eruptions based on spectral and image data. After discovering an outbreak, the software was pre-programmed to use this data and planned follow-up observations. For example, if the spacecraft detects a thermal emission in excess of two megawatts, the spacecraft should continue to observe it on the next overflight. The AI software aboard the spacecraft already knows when it will fly over the emission next, calculating how much space is required for observation on the semiconductor writer, as well as any other variables required for the next run. The software can also displace other observations from orbit to prioritize emerging science.
2020 and beyond
"This is a great example of things we could do and in the future become more complicated missions," said Chien. "Now we want to integrate a similar planning system into the Mars 2020 Rover, which is much more complicated, because a satellite follows a very predictable orbit, so the scientific data it collects have to deal only with variables."
"If you plan to At 10 o'clock, to take a picture of this volcano, take a picture of the volcano pretty much at 10 o'clock, because it is very easily predictable, "Chien continued." Unpredictable is whether the volcano erupts or not, so the AI will used to respond to it. "A rover, on the other hand, has to deal with a huge collection of environment variables that shift from moment to moment.
Even for a orbiting satellite, planning observations may be AI plays an important role, even when a human being makes the decisions, Chien said, "Depending on the mission complexity and the number of constraints that can be included in the software, this can The person can focus on priorities, see what different schedules come out, explore a larger part of the room to make better plans, or simply automate this for simpler missions.
Despite the lessons of EO-1, Chien said spacecraft using AI are "the exception, not the norm." I can tell you about various space missions that use AI, but if you happen to be using one Space mission, the likelihood that they are AI in a significant W used, very low. As a practitioner, we need to increase the intake. That will be a big change. "